Most of the components currently used in proton exchange membrane (PEM) fuel cells are derived from designs originally developed for use in phosphoric acid fuel cells (PAFC), and are not optimal for the higher performance of PEM fuel cells.
By the mid-1970s, components consisting entirely of carbon were made for use in PAFC's operating at temperatures in the 165-185.degree. C. range. In the case of one manufacturer (Energy Research Corporation, Danbury, Conn.) bipolar plates were molded from a mixture of graphite powder (approximately 67 wt %) and phenolic resin (approximately 33 wt %) and were carefully heat-treated to carbonize the resin without introducing excessive porosity by rapid degassing. Typically, heat treatment to 900.degree. C. was sufficient to give the required chemical, physical and mechanical properties. Initially bipolar plates were molded flat and were machined to produce the required reactant gas distribution grooves (or cooling grooves for the bipolar plate). Later, grooved plates were molded in a die (which was slightly oversized to compensate for shrinkage during baking) to produce the glassy graphitic, carbon-composite plate. In work performed at Westinghouse in the late 1970s/early 1980s the "straight through" gas distribution grooves on the bipolar plate were redesigned to yield an arrangement which has become known as the Z-plate.
The bipolar/separator plate in United Technologies Corporation's (UTC's) 1 MW demonstration stack (ca. 1975) was molded from graphite powder and polyphenylene sulfide resin. The corrosion resistance of this plate was shown to be only marginally acceptable in the finished demonstrator. Therefore, as shown in FIG. 1A, the plates in the 4.5 MW New York demonstrator 10 (ca. 1978) and also in subsequent UTC stacks were prepared by molding from graphite powder and inexpensive resins, followed by baking and graphitization at about 2700.degree. C. The surfaces of the molded ribs were then finished by sanding. In the 4.5 MW demonstrator 10, separator plates 12 lying between the anodes 14 and cathodes 16 of adjacent cells were ribbed on both sides to provide gas channels arranged perpendicular to each other.
As shown in FIG. 1B, the later 40 kW on-site units fabricated by UTC (ca. 1983) used a new ribbed substrate stack 20. This system placed the gas distribution channels 22 in the porous electrode substrate 24 itself, rather than in the flat bipolar plate 26, which was about 1 mm thick. The ribbed sides of the substrate 24 contacted the surface of this flat bipolar plate. The catalytic electrode mix 28 was applied to the opposite sides. The initially perceived advantage for this technology was reduced cost, since it offered the possibility of molded bipolar plates requiring a minimum of surface finishing, together with ribbed substrates of relatively low porosity that would be easy to machine. Pressurized PAFCs require the use of fully, or at least partially, graphitized bipolar plates and electrocatalyst substrates with heat treatment temperatures of at least 1800.degree. C. and preferably 2700.degree. C., or alternatively glassy carbons produced at high temperature.
The technology for making carbon/graphite bipolar plates for PAFCs has been used in PEM fuel cells by all the major PEM fuel cell developers (International Fuel Cells, Inc., Ballard Power Systems, H-Power Corp., Energy Partners, Fuji, and Siemens). While it is effective, it is expensive, and it is difficult to produce thin carbon based bipolar plates, and consequently stacks built with these plates tend to be heavy and bulky.
An obvious approach to overcoming these limitations is to use a moldable graphite-based composite that does not have to be carbonized. In this type of material graphite powder, which serves as the conductor, is bonded into a rigid piece with a polymer matrix. The graphite retains its conductivity and corrosion resistance, and the polymer binder, which must also be stable under PEM operating conditions, allows it to be formed by conventional polymer forming processes.
This approach was examined by General Electric in the early 1980's, and has been used successfully by Energy Partners for the 7 kW stack that they built for the Ford Motor Company as a vehicle prototype.
This approach has distinct limitations. When the graphite is diluted with the polymer, its conductivity, already lower than any metal useful in this application, is reduced even further. A seven kilowatt stack with pure graphite bipolar plates would be expected to have a 16 Watt internal resistive loss. When the graphite is dispersed in a polymer matrix, this loss will be larger. This is clearly shown by the data in Table I, which contains the properties of a number of materials potentially useful for fabricating bipolar plates for PEM fuel cell stacks. This table gives the specific resistivity and density of each element. Also included in the table are the mass and through plate resistance for the flow field region of a bipolar plate made from each material. This hypothetical plate was modeled as being 3.75 mm thick, with an active area of 125 cm.sup.2 and a serpentine flow field having channels 1.5 mm deep occupying 50% of the active area.
Table I Properties of Materials Useful for Bipolar Plates Specific Resistivity Element (.mu..OMEGA.-cm) Density (g/mL) Mass (g) R (.mu..OMEGA.) Cu 1.673 8.89 250 0.009 Al 2.655 2.70 75.9 0.014 Mg 4.450 1.74 48.9 0.024 Ti 42.0 4.50 127 0.23 C.sup.a 1375 2.25 63.3 7.4 .sup.a Graphite
The weight comparisons in Table I are based on plates which have identical dimensions and different compositions. In all cases, the thickness of the gas barrier is 0.75 mm at the thinnest point. For a graphite plate, this is very thin. Given the porosity of most graphitic materials, using a barrier this thin would require filling the pores with a sealant to produce a reliably gas-tight barrier. With any of the metals shown, this barrier could be made even thinner, with further gains in both size and weight.
Replacing graphite with any of the metals would increase both electrical and thermal conductivity significantly. Cu has the greatest conductivity of the metals listed in Table I, but it also has a high density. A solid metal stack would be quite heavy. Other strategies for bipolar plate construction have been developed to overcome some of the aforementioned limitations. As an example, recent efforts at Los Alamos National Laboratory have focused on the application of various expanded metal screens as flow fields. The screens are backed by a thin metal plate of the same material to create a bipolar plate configuration for use in a stack. An advantage of this system is the low material and manufacturing costs. However, one major disadvantage is the poor conductivity that results from multiple interfaces and the screen/plate point contacts.
Like much other PEM fuel cell technology, the basic electrode structures used in most PEM fuel cells are derived from phosphoric acid fuel cell (PAFC) technology. A conventional electrode structure 30, as shown in FIG. 2, has a thin layer of Pt 32 supported on high surface area carbon 34 as the active electrocatalyst. This is supported on a much thicker gas diffusion layer 36 typically consisting of an open matrix PTFE bonded carbon powder composite impregnated into a conductive carbon cloth support. The carbon support is in contact with the graphitic or metallic flow fields 38 on the bipolar plate. An alternative design uses conductive carbon paper to serve both the gas diffusion and support functions. A more recent variation in this design has an even thinner electrode, (described as either a thin layer electrode or an ink electrode) fabricated directly on the membrane 39. This electrode, while using less Pt, still uses the same gas diffusion structure as the conventional electrode.
All of these electrode structures are based on carbon technology. This puts some severe limitations on their performance. Not only is carbon a relatively poor electrical and thermal conductor, but the carbon-PTFE gas diffusion structure must be kept highly compressed to maintain a sufficient number of particle-to-particle contacts between carbon particles in the open polymer matrix to maintain its electrical conductivity.
Even the poorest of the metallic conductors has a resistivity over an order of magnitude lower than graphite. With the exception of magnesium, all of the metals shown are denser than graphite as well. While this would make a solid metal component quite heavy, this density is no bar to the inclusion of the metals in highly porous forms. Ni has better thermal conductivity than graphite, while Ti is somewhat poorer. This reduction in thermal conductivity for Ti is at least partly offset by the fact that Ti components can be made much thinner than the corresponding graphite ones, with a shorter path reducing the total resistance. The replacement of carbon/graphite based materials with metal components is expected to greatly enhance both the electrical and thermal conductivities as well.
The current practice of using thick gas diffusion layers forces a compromise between mass transfer capabilities and electrical resistance. Additionally, if the gas diffusion layer is compressed to enhance conductivity, the compression itself would further inhibit mass transfer. Replacing the carbon based structure with a metal structure that is both more porous and more conductive will simultaneously improve gas distribution, electrical conductivity, and thermal conductivity. Integrating the gas diffusion element into the structure of a lightweight metallic flow field will further reduce the weight of the stack while it improves both electrical and thermal conductivities, and opens up the possibility of a fully unitized structure with metallurgical bonds from the anode gas diffusion element of one cell to the cathode gas diffusion element of the next cell.
The necessity of keeping conventional carbon based gas diffusion structures compressed, including both PTFE bonded carbon powder and conductive carbon paper based gas diffusion structures and electrode assemblies that go with them, is well known in order to obtain good conductivity. It is one of the key reasons that most conventional PEM fuel cell stack designs use a heavy filter press arrangement 40, like that shown in FIG. 3, having heavy endplates 42, bolt holes 44 and cell frames or components 46 disposed between the endplates.
While the filter press arrangement is conceptually simple and easily executed, it has some distinct drawbacks. The largest of these drawbacks is the amount that it adds to the weight and the bulk of the stack. Since a large compressive force is required, the stack must have heavy tensile members (tie rods) to apply this force, with large terminations, and heavy end plates to distribute this force evenly over the area of the stack. Both of these features add to the weight and the bulk of the fuel cell stack. Also a problem, but not as widely recognized, is the poor thermal conductivity of porous structures based on polymer bonded powdered carbon and other porous carbon materials.
FIG. 4 shows an apparatus 50 that was used to ascertain the contribution of each of the materials in a standard fuel cell stack to the overall heat transfer. This device used an electrical heater 52 to heat a copper plate 54 to simulate the heat generated by a membrane and electrode assembly (M&E) during normal operation. A typical gas diffusion structure 56, type ELAT from E-Tek, Inc., Natick, Mass., was placed against the copper plate. This was backed by a flow field element, with two different types illustrated here, an expanded Ti sheet 58, at left, and a sheet of Ni foam 60, like that shown in FIG. 5, at right. The central component was a water cooled bipolar plate assembly 62.
To determine the thermal resistivity of the components, a stable cooling water flow is established and power is supplied to the heating elements. The temperature of the two heating elements 52 was raised until the copper plate 54 was at the desired M&E operating temperature and the unit was left in this state until a steady state was achieved, based on a stable cooling water exit temperature and no change in any of the readings of the thermocouples 64 (shown as dots).
In repeated tests, at a variety of copper plate temperatures between 50 and 85.degree. C., it was found that the face-to-face gradient of the ELAT gas diffusion element was greater than that of the metal flow fields. This was true even in the case of the Ni foam flow field 60, with 95% open volume and over five times the thickness. The thermal resistivity for the Ni foam was found to be 0.05.degree. C./Watt compared to 0.26.degree. C./Watt for the carbon powder, carbon cloth, PTFE composite gas diffuser. Clearly porous carbon gas diffusion structures based on conductive carbon cloth or paper are an impediment to better heat distribution and dissipation in a PEM fuel cell stack.